.json`. Ensure that the default settings look correct. These settings can be overridden using config generators ([see this tutorial](https://mimicgen.github.io/docs/tutorials/launching_several.html)).
+
+### Step 3: Execute Data Generation Workflow
+
+You are now all set to try data generation. You should be able to follow the steps documented in the [Data Generation Workflow Overview](https://mimicgen.github.io/docs/tutorials/getting_started.html#overview-of-typical-data-generation-workflow) with some minor changes.
+
+
+
Note
+
+Now that you have followed these steps, you can generate datasets for any other variants of this task, as long as the TaskSpec does not change -- e.g. different object placements, different robot arms, and different object instances. The [Reproducing Experiments](https://mimicgen.github.io/docs/tutorials/reproducing_experiments.html) tutorial provides examples of all three variations.
+
+
+
Note
+
+Before starting, you should ensure that a [robomimic environment wrapper](https://robomimic.github.io/docs/modules/environments.html) exists for the simulation framework you are using. See [this link](https://robomimic.github.io/docs/modules/environments.html#implement-an-environment-wrapper) for guidance on how to create one. The environment metadata in the source hdf5 should point to this environment wrapper.
+
+
+
Warning
+
+When you change the `get_subtask_term_signals` method, you should re-run the `prepare_src_dataset.py` script on the source data to re-write the subtask termination signals to the dataset.
+
+
+
Note
+
+This section helps users get started with data generation. If you would just like to download our existing datasets and use them with policy learning methods please see the [Reproducing Experiments](https://mimicgen.github.io/docs/tutorials/reproducing_experiments.html) tutorial for a guide, or the [Datasets](https://mimicgen.github.io/docs/datasets/mimicgen_corl_2023.html) page to get details on the datasets.
+
+
+
Warning
+
+If you do not edit `scripts/generate_core_configs.py` the default settings will run data generation until 1000 success trajectories have been collected. This is why it is important to set `NUM_TRAJ = 10` and `GUARANTEE = False` for a quick run. Alternatively, pass the `--debug` flag to the command in Step 3, which will run an even smaller data generation run.
+
+
+
Note
+
+If you run into a `RuntimeError: No ffmpeg exe could be found.` at the end of the script, this means rendering the dataset to video failed. We found that a simple `conda install ffmpeg` fixed the problem on our end (as documented on the [troubleshooting page](https://mimicgen.github.io/docs/miscellaneous/troubleshooting.html)).
+
+
+
Note
+
+The generated `demo.hdf5` file is fully compatible with robomimic, which makes it easy to inspect the data (see [here](https://robomimic.github.io/docs/tutorials/dataset_contents.html)) or launch training jobs (we have a helper script at `scripts/generate_core_training_configs.py`).
+
+
+
Note
+
+You should ensure that a [robomimic environment wrapper](https://robomimic.github.io/docs/modules/environments.html) exists for the simulation framework you are using. See [this link](https://robomimic.github.io/docs/modules/environments.html#implement-an-environment-wrapper) for guidance on how to create one. The environment metadata in the source hdf5 should point to this environment wrapper.
+
+
+
Note
+
+See the [Data Generation for Custom Environments](https://mimicgen.github.io/docs/tutorials/datagen_custom.html) tutorial for a more comprehensive description of implementing environment interfaces for new simulators and new tasks.
+
+
+
Note
+
+Data generation does not necessarily need to run on the exact same task. As described in the paper, you can run it on tasks with possibly different reset distributions, different robot arms, or different object instances. The [Reproducing Experiments](https://mimicgen.github.io/docs/tutorials/reproducing_experiments.html) tutorial provides examples of all three variations.
+
+
+
Note
+
+We recommend going through the [Getting Started](https://mimicgen.github.io/docs/tutorials/getting_started.html) tutorial first, so that you are familiar with the way data generation works.
+
+
+
Note
+
+This script requires the `pygame` module to be installed (it will not be installed by default when installing the Mimicgen repo).
+
+
+
Note
+
+We provide some utilities to debug your choice of subtask termination signals and offsets -- see the [Debugging Data Generation](https://mimicgen.github.io/docs/tutorials/debugging_datagen.html) tutorial for more information.
+
+
+
Note
+
+You can find task reset visualizations on the [website](https://mimicgen.github.io), but they may look a little different as they were generated with robosuite v1.2.
+
+
+ 
+